Machine Learning-Based Model in Predicting the Plate-End Debonding of FRP-Strengthened RC Beams in Flexure

Abstract

Reinforced concrete (RC) beams strengthened with fiber reinforced polymers (FRPs) are structurally complex and prone to plate-end (PE) debonding. In this study, considering the extremely complicated nonlinear relationship between the PE debonding and the parameters, machine learning algorithms, namely, linear regression, ridge regression, decision tree, random forest, and neural network improved by sparrow search algorithm, are established to predict the PE debonding of RC beams strengthened with FRP. The results of reliability evaluation and parameter analysis reveal that ACI, CNR, fib-1, fib-2, and TR55-2 are a little conservative; AS and TR55-1 have the problem of overestimating the shear force; the accuracy and robustness of the SSA-BP model developed in this paper are good; the stirrup reinforcement has the greatest effect on PE debonding; and each parameter shows a complex nonlinear relationship with the shear force when PE debonding occurs.